Methods for separating and detecting a transmission stream on the receiver side in a spatial multiplexing type MIMO communication for transmitting different information from multiple transmission antenna include methods using a spatial filter such as a ZF (Zero Forcing) method and an MMSE (Minimum Mean Square Error) method as in Non-Patent Document 1. However, sufficient performance cannot be achieved in these methods, though the computation is not complicated.
Meanwhile, MLD (Maximum Likelihood Detection) demodulation for selecting the closest signal points from among all the signal points that can be taken by the transmission signals as a result of the estimation of a transmission path is also known. Though this has high detection performance, the computation scale increases exponentially when the number of antennas or the number of modulation multiple values increase, which makes it difficult to be applied to an actual system.
In recent years, super multi-level modulation systems such as 256 QAM (Quadrature Amplitude Modulation) and 1024 QAM have been adopted as a wireless communication system. In contrast to these, it is obvious that the application of MLD is not realistic.
A typical method for decreasing the computation scale in an MLD is an SD (Sphere Decoding) method. According to an SD method, reception candidate points within a certain distance range from the reception point are narrowed down from among the generated reception candidate points (replicas) so that an MLD operation is carried out targeting the narrowed-down reception candidate points.
In Patent Document 1, a hard decision is made on the results of demodulation gained in accordance with a ZF method or an MMSE method, and the hard determination points and the points in its vicinity are used as transmission candidate points in order to reduce the MLD operation by reducing the transmission candidate points and the reception candidate points that are generated therefrom.